Lock one decision window
Choose a review point: the end of the current war, arrival of another enemy army or the moment treasury must regain a target reserve. Both alternatives need the same number of turns.
Loot enters treasury now and can become reinforcement or urgent infrastructure. Holding creates future flow while locking capital into recovery and defence first.
Count only realizable provincial flow
Record income after recovery and subtract the additional commitments created by the frontier. If another army is required to defend it, that upkeep belongs to the cost of holding rather than an abstract war budget.
Do not pull distant income beyond the review point into the case simply because the province could theoretically exist forever. The campaign pays for liquidity and tempo now, not an unreachable infinite-horizon total.
Add loss risk and the cost of another front
Risk does not need false precision. Compare a baseline with a hard case: lower income, longer recovery and loss before the horizon ends. The difference exposes decision sensitivity.
Patch 8.1 makes the test more important in late campaigns. More active AI priorities increase the value of response liquidity without creating one universal answer for every border.
Separate the financial result from the strategic objective
The model may prefer sacking on gold while access, a resource or removal of a threat still makes occupation correct. Record that strategic reason separately instead of hiding it inside invented ROI.
The final note should state the choice, horizon, required reserve and review trigger. The decision remains explainable several sessions later.
Replace the baseline with your inputs and get an answer for your scenario.
Data note. Every numeric example is a transparent baseline scenario. It is not presented as official publisher data and can be recalculated in the linked tool.